A model is presented for analyzing the components of instructional programs that are likely to affect generalization with severely handicapped students. The literature is summarized in terms of such issues as natural vs. artificial environments, materials, cues, and contingencies. The authors also review empirical evidence about the relative efficacy of serial vs. concurrent presentation, prompts vs. corrections, and various instructional arrangements. Suggestions are offered for implementation, and research needs are described for several components. Future research on the interactive effects of the components with each other and with the learners' characteristics are discussed.